A follow-up to the original framework post documenting the refinements introduced this week. The indicator set, report structure, and adaptive methodology are unchanged — what follows is a summary of the precision and consistency improvements made to the scoring and labelling logic.

Summary of Changes

Area What changed
Cross annotations The “Death Cross” / “Golden Cross” label is now reserved strictly for the classic SMA 50/200 pair. Any other pair — such as SMA 20/50 in a short window — is labelled “Bearish Cross (SMA20/SMA50)” instead, avoiding overstating the signal’s severity.
Signal Summary table The row label, the SMA pair compared, and the Bollinger Width threshold all now adapt to N_DAYS, matching the indicator choices shown in the charts.
Volatility scoring BB width above 3% now reduces vol_score by 0.5; above 6% by 1.0. Previously the prose described “elevated” volatility while the score remained 0, making the badge and the weighted verdict inconsistent.
Trend conclusions Section 1 of the conclusions engine now uses the same window-adaptive SMA pair shown in the charts, adds a gap-widening check for that pair, and always evaluates the SMA 50/200 Death Cross as a fixed institutional reference regardless of window size.
Momentum signal labels Two new signal states — “Mixed / Caution” and “Oversold / Caution” — prevent overbought oscillators combined with a bullish MACD from being mislabelled as “Bearish”. The section score is also capped at −0.15 in the Mixed / Caution case.
Verdict descriptions The Bullish and Bearish verdict text now has three confidence-adaptive variants (High / Moderate / Low) rather than one generic string, so a Low-confidence verdict no longer reads as more certain than it is.
Verdict page layout Score pills now display each section’s signal label and are ordered left-to-right to correspond directly to table rows top-to-bottom. Column headers were revised and an explanatory sub-caption was added.
Ticker input input() prompts replaced with hardcoded TICKER and N_DAYS variables for Papermill compatibility.

Generated with Python and matplotlib · @lbarqueira.bsky.social